如何基于Pandas DataFrame统计各问题的回答占比？

首先还原你提供的DataFrame：

import pandas as pd

# 构造示例数据
data = {
    "Do you smoke?": ["yes", "no", "sometimes", "no", "no", "yes"],
    "Do you exercise?": ["no", "yes", "yes", "yes", "no", "no"],
    "Do you sleep less than 8 hours?": ["sometimes", "no", "yes", "sometimes", "no", "yes"]
}
df = pd.DataFrame(data)

方法一：直接生成宽格式占比表 #

对DataFrame的每一列使用value_counts(normalize=True)，直接得到该列各选项的占比，再转成百分比格式：

# 计算每列各选项的占比，转成百分比并保留两位小数
percentage_result = df.apply(
    lambda col: col.value_counts(normalize=True).mul(100).round(2)
)

print(percentage_result)

运行后输出结果：

Do you smoke?  Do you exercise?  Do you sleep less than 8 hours?
no                   50.0              33.33                            33.33
yes                  33.33             66.67                            33.33
sometimes            16.67              0.00                            33.33

方法二：生成规整的长格式占比表 #

如果需要更适合后续分析或可视化的长表格式（问题、选项、占比三列），可以先将数据转成长格式，再分组计算：

# 将宽表转成长表，提取问题和对应的回答
melted_df = df.melt(var_name="Question", value_name="Answer")

# 按问题和回答分组，计算每组占该问题总人数的百分比
long_format_result = (
    melted_df.groupby(["Question", "Answer"])
    .size()
    .groupby(level=0)
    .apply(lambda x: x / x.sum() * 100)
    .round(2)
    .reset_index(name="Percentage")
)

print(long_format_result)

运行后输出结果：

Question     Answer  Percentage
0                      Do you exercise?         no        33.33
1                      Do you exercise?        yes        66.67
2                         Do you smoke?         no        50.00
3                         Do you smoke?  sometimes        16.67
4                         Do you smoke?        yes        33.33
5  Do you sleep less than 8 hours?         no        33.33
6  Do you sleep less than 8 hours?  sometimes        33.33
7  Do you sleep less than 8 hours?        yes        33.33

关键说明 #

• value_counts(normalize=True)：返回各选项在列中的占比（0-1之间的小数）
• mul(100)：将小数占比转换为百分比
• round(2)：保留两位小数，让结果更易读
• melt()：将宽格式数据转换为长格式，方便按问题分组统计

内容的提问来源于stack exchange，提问作者JFCastello